Sorry if its more long winded stuff. It may be of interest to some. The following link maybe a good introduction to methods some shooters use.
https://www.youtube.com/watch?v=oKxf-dbexMM . It has a link to an Excel program, but beware I could only open it on an older version of Excel. 2019 said it was corrupt.
The author makes the point the concepts are
basic and they are, if you have a statistical background. Therefore, the holy grail some extol with low SD's and ES depends on the sample size. In isolation, five shots or even three do not have much confidence associated with them unless they have already come from a tuned rifle. Thirty is a minimum sample size to be statistically significant to determine a guide for tune. It needs to be repeated.
Before I examine the merits of statistical significance in a more advanced way for some that may be interested, there are two huge trade off's shooters have to bear in mind. Firstly, too many shots will ruin a good barrel. It’s probably at its peak around 400 rounds. By the time you break it in properly to get past the velocity hump of conditioning, the window for top draw accuracy becomes narrow. It should not be wasted. The second thing concerns further throat erosion for the period of top draw accuracy. Use time before the hump to condition cases. You should use a reduced load and condition the barrel at the same time then visit formal load development between 100 and 150 rounds IMO.
The initial goal of some shooters is to identify a suitable plateau in velocity by incremental load testing (.2 grains depending on case) with the use of a chronograph using safe loads above and below a pet load, watching for pressure signs on the upper limit. The plateau may contain an OCW pattern, a node or a compensation point or mix of all three.
Chronographs are valuable but, in the past, had a significant error factor. I use Quick load as the go to simulate pressures, barrel time etc. and the chronograph to confirm that. I measure radial spread from 140-yard target having optimized trade-offs with coning, wind and light effects. I only use ladder testing if applicable at very long range because there has been very different terminal flight characteristics in play as confirmed by Jim Boatright which I observed and documented too but could not explain at the time that can alter group shape. Albert may have experienced this at 1000 yards. I do record velocities using a Lab Radar at the muzzle and another Canadian kit at the target for confirmation.
Shooters usually refine charge with seating depth tests or then maybe neck tension to identify tighter groups. I look for stability as a priority, and not necessarily the tightest group. Within the range of seating depths, advanced consideration should be given to where expected erosion may take you. Hopefully a long node may assist and it is not so much an issue. The erosion rate is different between bullets jammed in the lands, just off the lands or well back. It is hard to follow the throat when matches are shot over a few days and in long strings by decreasing seating depth at the venue. That skinny node can tip you out if it’s not taken into account. So, a bit of homework should be done to establish benchmarks for erosion. A trade-off has to be made with the barrel lift process too.
Coming back to statistical significance, there is a big difference between correlation (which shows relationships) and causation. The random, or the outlier maybe discounted as noise etc. in correlation but it will have a cause. It may be a compensation factor, bum shot or a mirage error when sighting in short range load development. A good setup minimizes errors in the process. Much knowledge has been gained in the past from firing from a machine rest with a barreled action that had been properly put together. With many thousands of shots fired, I know if I am in the ball park with charge weight tests with 30 shots, in much the same way pet loads or good loads referred by others because they reflect significant numbers. So, I feel I can use less data to populate my tests, and particularly when the rifle has already been tuned and I am chasing the throat. You know when to stop with experience.
I am in the camp of ES these days for compensation tuning. It’s that so-called random shot or outlier that can lose a match. I mitigate against that to a point using harmonic theory and practice which I am sorry to say is an advanced craft to optimize. There is a third inertia point which I have referred to before long ago in posts, in the atmosphere before the bullet takes free flight which impacts on its exit from the muzzle which a forward of the muzzle tuner can influence. From a statistical point of view, the bell curve used in the video does not always represent a nice neat curve based on the other inertia points i.e. ignition and throat engagement of the projectile. The curve can be skewed left or right or in statistical terms represent kurtosis of high or flat peaks which could have fat tails or long ones. The distribution is distorted. Statisticians can verify by other means. Common tests may include chi square, "F" test and "T" tests if others want to check them out.
There are tests which give statistical validity to ES. Again, I use the actual load development target from 140 yards and do not rely solely on the chronograph but the actual radial measurement of the spread. I think it was a statistician by the name of Grubb using a Monte Carlo simulation with a population of 1000 data points who determined a formula which indicated how much population was required for an error factor reduced to 10%. If this figure is applied as a minimum number of shots to assess ES, it comes out at 7 shots. If several strings of seven are used with incremental load then that error reduces, but you have to shoot quite a few more to be sure. I can live with 10% error with the use of ES as a discriminator.
I don't necessarily rely of linear regression, but do use nonlinear regression to fit curves such as ES data. I have posted in detail about this process before. This post becomes too long to repeat that information and images in Photobucket have gone to God on my old posts.
